The Link Between Roll and Steering Torque inMotorcycles

Thomas Lane and Roberto Lot

Abstract—Steering torque is the most important control inputfor motorcycle turning; however, some riders believe they cansteer using upper body roll alone. As other investigations haveshown rider roll to be an ineffective for path control, thereis likely a link between rider roll and steering torque inmotorcycles. This paper examines new results, looking at therelationship between roll and steering. The specific results lookat the connection in two directions. First, results examine howsteering torque is generated due to motorcycle roll. Secondly,the rider’s roll and yaw response to steering motion is presented.The results show strong links between motorcycle roll rate andsteering, as well as steering and rider motion. The link betweensteering and rider roll implies the rider can generate steeringtorque by leaning their upper body. As the generated forces andmotions are non-negligible they need to be investigated further.This information, is being used to develop a new model of therider’s passive impedance. In turn this model can be used toensure the stability of motorcycles, regardless of the rider.

Index Terms—motorcycle, rider, roll, steering-torque, control

I. INTRODUCTION

ONE of the most important controls for a motorcyclistis steering torque. Steering torque can be intentionally

provided by the rider or caused by other motions, such asrider roll. Understanding how steering torque is generated isof key importance, as it is the primary method of directionalcontrol for a motorcycle.

A common concept in motorcycle riding is that a rider cancorner by leaning their body into the turn, rather than usingsteering torque. This idea is refuted by [1] that shows steeringtorque is more than ten times as effective, at turning amotorcycle, than rider roll alone. This idea is shown directlyby [2], meaning rider roll must have a secondary effect. Thispaper intends to examine the possibility of a link betweensteering torque and roll, of both motorcycle and rider.

This paper is part of a larger investigation on how therider affects motorcycle stability. The rider can respond tomotorcycle motions in two ways, actively or passively. Theactive response occurs at frequencies below 2 Hz, where therider has control of their body. This active response includesfunctions such as keeping the trunk vertical, and controllingsteering inputs. Above 2 Hz [3] the rider can no longercontrol their body to respond to motions in time, this isthe passive region. In the passive region the rider has no

Manuscript received March 06, 2018; revised March 23, 2018. Thisresearch was funded by the EPSRC. The data created by this investigationwill be made available through the University of Southampton, once thisresearch has been completed and the related thesis published.

Thomas Lane is with the Department of Mechanical Engineer-ing, University of Southampton, Hampshire, SO17 1BJ, UK e-mail:TL14G15@soton.ac.uk.

Roberto Lot is with the Department of Mechanical Engineer-ing, University of Southampton, Hampshire, SO17 1BJ, UK e-mail:Roberto.Lot@soton.ac.uk.

cognitive effect on body movement, so can be modelled asa multibody system of masses, springs, and dampers.

When exploring the link between roll and steering, itis useful to include this passive region, as the rider canbe measured without knowingly changing the result. In theactive region, if the rider knew the purpose of the test, theymay change how they use their arms or trunk, affecting thesteering torque result. By using a frequency that the ridercan’t actively respond to, the result is likely to be closer tothe real world case, where they are focused on path followingand navigation.

The order of the paper is as follows: first there is anintroduction to the equipment used for this investigation,followed by an explanation of the tests undertaken. Resultsare presented to show the effect of motorcycle roll on steeringtorque, as well as steering movement on rider roll and yaw.These results are analysed to understand the possible linkbetween rider roll and steering torque, along with suggestionsfor further research.

II. SPECIALIST EQUIPMENT

This investigation used a rig specially designed to measuremotorcycle rider impedance, pictured in Figure 1, to ensurethe results represented the real world. This rig is capableof replicating motorcycle roll, yaw and steering motions,with the rider’s posture matched closely to currently availablemotorcycles. This rig also allows the posture to be altered,representing either sports or touring motorcycles for the firsttime, depending on the test requirements. An important dif-ference between this rig and used in previous investigations,is the location of the roll axis. Whilst [4] & [5] use a rollaxis along the ground, this rig places the roll axis about ∼100mm below the seat. This location is thought to give a betterrepresentation of motorcycle weave and wobble.

This rig uses a bespoke measurement system that collects34 channels of data. This system consists of 4 six axis inertialmeasurement units (IMUs) and an assortment of load cells.Two IMUs record rig motion, one being placed to measureroll and yaw rotation, whilst the other measures steeringrotation. The other two IMUs are attached to the rider’supper and lower back, aligned with the spine, to record ridermotion.

Bespoke load cells are used with this system, measuringhow the rider interacts with the motorcycle. Two types ofcell are used. The first measures lateral force and momentapplied to the seat, tank, and foot pegs. The second type isused for the handlebars only, measuring roll torque, verticalforce, lateral force and steering torque. A key benefit ofthis system, over those from previous investigations, is thatevery measurement is taken on every test. This technique isallowing new connections to be explored, such as the link

Proceedings of the World Congress on Engineering 2018 Vol II WCE 2018, July 4-6, 2018, London, U.K.

ISBN: 978-988-14048-9-3 ISSN: 2078-0958 (Print); ISSN: 2078-0966 (Online)

WCE 2018

Fig. 1. Image of the motorcycle rig used for testing. Roll frame in grey,Yaw frame in black, and handlebar in silver. Rotation axes are highlightedin white

between rider roll and steering torque. The full capabilitiesof this rig, as well as more details about the design andoperation, are available in [6].

III. TEST PROTOCOL

Rider forces and motions can be classified as either activeor passive. Active inputs are those which the rider intendsto make, whereas passive inputs are those made by the riderinvoluntarily. These passive inputs can either be generatedas a response to an intended motion, or as a reaction toan external input. In the active case the rider may changetheir behaviour dependent on where their attention is placed.For instance, if a rider knows that the steering torque isbeing measured, being conscious of this fact may changethe way they use their arms if asked to lean their torso.As such, using an input they can only respond to passively,and focusing their attention elsewhere, is most likely to giveresults that represent the real world. To ensure the rider hasa full passive response, a vibration input can be used. Thisworks as humans can only actively respond to vibrationsbelow 2 Hz [3]. Using this knowledge, a frequency sweepof 0.5 - 12 Hz was used for these experiments, allowingfor a comparison between active and passive response. Thisfrequency range was selected based on those used in previousinvestigations (such as [4] [5] [7]), which cover the range ofthe most dangerous motorcycle stability modes.

The other consideration was how to excite the rider. Themost straightforward approach would have been to move therider’s upper body directly; this was not feasible as it doesnot represent a normal input for the rider, which could skewthe results, as well as presenting an ethical issue. Anotherapproach would have been to excite the steering column witha given steering torque and measure the rider’s upper bodyresponse. This was discounted, as the current set-up is notcapable of this type of control. Therefore, to elicit the mostnatural response from the rider, an input that mimics normalriding is used; in this case steering and motorcycle roll inputs.The steering input uses a constant 1.75◦ rotation amplitudethroughout the frequency range. A constant amplitude waschosen as it overcomes limitations in shaker control whilstbeing acceptably comfortable for the rider at high frequen-cies. The limitation of this approach is that low frequencies

100 101

Frequency (Hz)

0

2

4

6

8

10

12

Ste

erin

g T

orqu

e (N

m)

Fig. 3. Overlay plot showing the results of steering torque for 8 ridersduring a motorcycle roll test. The average result is shown in black

have a relatively low input energy. To help spread the inputenergy an exponential swept sine is used, which changesfrequency faster at higher frequencies. Whilst the frequencychanges in the same manner, the amplitude is different forthe roll tests. A constant velocity amplitude, with a maximumrotation of 6◦ at 0.5 Hz was chosen. Using a constant velocityamplitude provides a good level of input energy throughoutthe frequency range. This input also gives accessible rollangles at the start of the test, and tolerable acceleration levelsat the end.

In the context of this paper the key measurements aresteering and roll rate for the rig. The key rider measurementsare steering torque and rider upper body roll rate. Thesemeasurements were selected from those available, as theyallow the relationship between the rider and steering to beinvestigated in two directions. Motorcycle roll and steeringtorque allow for a direct excitation of the rider, but it isnot certain if rider roll is directly causing the output. Thisresult can be sided by exciting the steering rotation to see theeffect on rider body roll. From these quantities it is possibleto infer, if not calculate, the relationship between rider rolland steering torque.

IV. RESULTS

A. Steering torque due to motorcycle roll

The first result to be examined, is the relationship betweenmotorcycle roll and steering torque. This is a good startingpoint, as it is the best alternative to exciting the rider’s upperbody directly.

The first consideration is the magnitude of steering torque;no further investigation would be necessary if the levelswere low. Figure 2 shows a magnitude of approximately 10Nms/radian, throughout the frequency range. This equatesto an average steering torque of 2 - 4 Nm below 6 Hz, asshown by Figure 3. This torque level is significant, as therider can use 8 Nm to generate 20◦ roll angles during aslalom manoeuvre [8].

Secondly, to ensure this torque is being generated inresponse to roll, rather than at random, the coherence must be

Proceedings of the World Congress on Engineering 2018 Vol II WCE 2018, July 4-6, 2018, London, U.K.

ISBN: 978-988-14048-9-3 ISSN: 2078-0958 (Print); ISSN: 2078-0966 (Online)

WCE 2018

100 101-20

0

20

40M

agni

tude

(dB

)

100 101

-200

0

200

Pha

se (

deg)

100 101

Frequency (Hz)

0

0.5

1

Coh

eren

ce

Fig. 2. Overlay plot showing the results of steering torque / Rig roll rate for 8 riders. The average result is shown in black

checked. Figure 2 shows high coherence across the frequencyrange for most riders, increasing as the rider’s responsebecomes more passive. This is the first investigation to showthat steering torque is generated due to motion in anotherplane. The magnitude of the steering torque, indicates thisconnection should not be omitted from motorcycle stabilityinvestigations, or the errors could be significant. Therefore,how this steering torque develops must be investigated.

Finally, we must establish if the rider causes this steeringtorque, or if it is due to some other factor. As the steeringcolumn is inclined, it could be created due to the roll motionexciting the steering directly, with the handlebar inertiagenerating the torque. However, there are two issues withthis hypothesis. In this experiment the force would increaselinearly throughout the frequency range, as acceleration does;The results are much more constant throughout the frequencyrange and drop off at higher frequencies. Secondly, theform of steering torque generated, and the magnitude, differssignificantly from that obtained in preliminary tests withouta rider. The rider is the only other connection to the steeringcolumn, and the only one flexible enough to generate theobserved response. As such it must be concluded that therider generates this steering torque.

Now that the rider is shown to cause of this steering torque,the link between rider and steering must be examined moreclosely. Specifically, the rider must be excited by the steering,and their upper body response measured.

B. Rider roll and yaw due to steering input

The most direct way of testing the link between rider’storso and steering, using the current equipment, is to measurethe rider’s motion as a result of steering movement. Riderroll is presented first as a strong connection would showgood evidence for our hypothesis. Rider yaw is also shownfor comparison, as it is the most directly excited by steering

Figure 4 shows how the rider rolls in response to thesteering input. A strong link between steering and rider rollis seen above 3 Hz, where the coherence is around 0.96. Thecoherence could be caused by the rider moving of their ownaccord, or due to the low input energy at lower frequencies.The reason is likely to be a combination of these two factors.

The motion ratio varies between 0.1 & 0.4, showing thelink between rider roll and steering rate. This result wouldindicate a motion ratio of 10 to 1 between rider roll andsteering, if it could be inverted. This is not directly possibleas in the real world the rider leans far to much to assumelinearity; the coupled dynamics on a motorcycle also meanthat high steering rates would be hard to achieve. However,this result does show steering torque could be generated dueto rider roll.

The result for upper body yaw to steering, also shows astrong connection (Figure 5). Firstly, the coherence is betterthan for roll in the low frequency range, showing yaw ismore reliant on the steering input than roll in this range.In part, this could be due to the rider not needing to rotatetheir spine for balance. The yaw response to steering is alsolarger than roll, with a peak of 0.8. This is expected as thetorso is more closely aligned with the steering column. This

Proceedings of the World Congress on Engineering 2018 Vol II WCE 2018, July 4-6, 2018, London, U.K.

ISBN: 978-988-14048-9-3 ISSN: 2078-0958 (Print); ISSN: 2078-0966 (Online)

WCE 2018

100 101-40

-20

0M

agni

tude

(dB

)

100 101

-400

-200

0

200

Pha

se (

deg)

100 101

Frequency (Hz)

0

0.5

1

Coh

eren

ce

Fig. 4. Overlay plot showing the results of rider upper body roll rate / steering rotation for 8 riders. The average result is shown in black

100 101

Frequency (Hz)

0

2

4

6

8

10

Ste

erin

g T

orqu

e (N

m)

Fig. 6. Overlay plot showing the results of steering torque for 8 ridersduring a motorcycle steering test The average result is shown in black

is represented in the result, as the peak rider response is thecosine of angle between rider’s torso and steering column,which is around 40◦ in this test.

A key difference between roll and yaw results is the fre-quency of maximum response. Yaw is greatest at 4 Hz, whilstroll is greatest around 7 Hz. This result shows that whilst thesteering input generates mainly yaw at low frequencies, asexpected, roll is dominant above 6 Hz.

Another point of interest is how the steering input gen-

erates steering torque. Figure 6 shows a similar trend insteering torque to the coherence of rider roll and yaw. Forevery participant except one, there is very little steeringtorque below 2 Hz, this indicates that the rider is relaxed andtrying not to impede steering motion. The torque generatedhere is due to acceleration the handlebar and portions of therider’s arm. As frequency increases into the passive region,the steering torque increases, as the rider can no longercontrol their arms well enough to isolate their upper body;The increased effective mass increases the steering torque.

V. HOW STEERING TORQUE IS GENERATED FROM RIDERROLL

The following is a hypothesis as to how steering torque isgenerated from rider roll. As the rider leans on the motor-cycle their shoulders move laterally relative to the steeringcolumn. This movement causes their arms to elongate androtate across their body. To retain a more neutral positionthe rider turns their shoulders towards the steering column,this movement makes their arms an uneven length. To matchthe length of each arm, so it does not feel like they aresteering, the rider attempts to turn the steering parallel totheir shoulders. Trying to straighten their arms has thusgenerated the steering torque.

This hypothesis should also be valid in the vibration case,as the rider’s body is acting as a series of springs anddampers, leading to a similar effect when the vibration issmall. Further work on modelling the rider, specifically theirarms, will give further insight into how the steering torqueis generated.

Proceedings of the World Congress on Engineering 2018 Vol II WCE 2018, July 4-6, 2018, London, U.K.

ISBN: 978-988-14048-9-3 ISSN: 2078-0958 (Print); ISSN: 2078-0966 (Online)

WCE 2018

100 101-40

-20

0M

agni

tude

(dB

)

100 101-400

-200

0

Pha

se (

deg)

100 101

Frequency (Hz)

0

0.5

1

Coh

eran

ce

Fig. 5. Overlay plot showing the results of rider upper body yaw rate / steering rotation for 8 riders. The average result is shown in black

VI. CONCLUSION

This paper has shown several important links between themotorcycle, rider, and steering system. A direct link betweenmotorcycle roll and steering torque has been shown. Thislink creates steering torque of a non-negligible magnitude,with good coherence throughout the frequency range. Thisoutput must be a product of the rider, as the output issignificantly larger with a participant on the rig. Strong linksare also seen between steering input, roll, and yaw of therider’s upper body. Coherence is good for both FRFs above2 Hz, where the rider’s response becomes passive. There isa greater magnitude of yaw below 6 Hz, whilst rider rollis more prominent above. The FRFs indicate that it wouldbe possible for the rider to create steering torque purely byrolling, though further investigation is needed.

To prove this hypothesis, further tests should be conducted.It would be possible to use static tests, where rider is asked tolean and steering torque is measured. The directness of thisapproach is an advantage, but results may be skewed as therider is aware of the purpose of this test. This knowledgemay cause the rider to react differently than normal, thusskewing the results.

This paper is part of a larger work on motorcycle stabilitybeing undertaken at the University of Southampton. Thisproject aims to create a comprehensive model of the rider’spassive response, based on new data. The more comprehen-sive data set, including links between motions and forcesin different planes, will allow the rider’s response to berepresented more faithfully. The final model will providegreater insight into the rider’s effect on motorcycle stability,

including how steering torque is generated due to roll.

REFERENCES

[1] R. S. Sharp, “Stability, control and steering responses of motorcycles,”Vehicle system dynamics, vol. 35, no. 4-5, pp. 291–318, 2001.

[2] K. Code, “A twist of the wrist ii [dvd],” 2009.[3] M. Massaro and D. Cole, “Neuromuscular-steering dynamics: motorcy-

cle riders vs. car drivers,” in ASME 2012 5th Annual Dynamic Systemsand Control Conference joint with the JSME 2012 11th Motion andVibration Conference. American Society of Mechanical Engineers,2012, pp. 217–224.

[4] T. Nishimi, A. Aoki, and T. Katayama, “Analysis of straight runningstability of motorcycles,” SAE Technical Paper, Tech. Rep., 1985.

[5] A. Doria, M. Tognazzo, and V. Cossalter, “The response of the rider’sbody to roll oscillations of two wheeled vehicles; experimental tests andbiomechanical models,” Proceedings of the Institution of MechanicalEngineers, Part D: Journal of Automobile Engineering, vol. 227, no. 4,pp. 561–576, 2013.

[6] R. Lot and T. Lane, “A new test rig for motorcycle riderimpedance measurement,” October 2017. [Online]. Available:https://eprints.soton.ac.uk/416217/

[7] V. Cossalter, A. Doria, R. Lot, and M. Massaro, “The effect of rider’spassive steering impedance on motorcycle stability: identification andanalysis,” Meccanica, vol. 46, no. 2, pp. 279–292, 2011.

[8] F. Cheli, P. Mazzoleni, M. Pezzola, E. Ruspini, and E. Zappa, “Vision-based measuring system for rider’s pose estimation during motorcycleriding,” Mechanical Systems and Signal Processing, vol. 38, no. 2, pp.399–410, 2013.

Proceedings of the World Congress on Engineering 2018 Vol II WCE 2018, July 4-6, 2018, London, U.K.

ISBN: 978-988-14048-9-3 ISSN: 2078-0958 (Print); ISSN: 2078-0966 (Online)

WCE 2018